Importance of Feature Selection for Recurrent Neural Network Based Forecasting of Building Thermal Comfort

The paper demonstrates the importance of feature selection for recurrent neural network applied to problem of one hour ahead forecasting of thermal comfort for office building heated by gas. Although the accuracy of the forecasting is similar for both the feed-forward and the recurrent network, the removal of features leads to accuracy reduction much earlier for the feed-forward network. The recurrent network can perform well even with less than 50% of features. This brings significant benefits in scenarios, where the neural network is used as a blackbox model of thermal comfort, which is called by an optimizer that minimizes the deviance from a target value. The reduction of input dimensionality can lead to reduction of costs related to measurement equipment, data transfer and also computational demands of optimization.